Background and Objective: Diagnosis of enamel caries and evaluation of proximal caries depth are some of the main problems in caries detection. In this study, a new method based on image processing techniques proposed that applied to radiographic images. The purpose of this method was to automatically identify and segment decayed areas of the teeth.
Materials and Methods: For this study which was done in Hamedan province of Iran, several molar and pre-molar teeth images with 158 inter-proximal surfaces were selected. Sixty teeth with dentinal caries, 31 enamel restricted caries, 11 DEJ limited caries, and finally 56 surfaces without any caries were detected. The teeth were placed inside the prepared acrylic blocks, and the intra-proximal and occlusal contacts of the teeth were reconstructed. Then digital imaging was performed, and the results considered as the input of the proposed image processing method. In the proposed method, after preprocessing of images, by applying morphological operators, and with the help of the K-means clustering, caries were automatically detected. The final results and information were analyzed by SPSS software.
Results: The sensitivity of the proposed method was 86.7% for dentinal caries, 67.7% for enamel caries, and 63.6% of caries that limited to DEJ. The specificity was 100%, and the overall sensitivity reported as 78.4%. There was no false positive, the false negative and the accuracy of the method were 21.6% and 86% respectively.
Conclusions: The findings of the present study show that the software designed to detect proximal caries has acceptable diagnostic accuracy. In the proposed method, based on observations, the sensitivity of caries detection improves with an increase in depth of decay. Therefore, it qualifies better in detecting dentinal caries than enamel and DEJ caries.